Adaptive learning and expansion of spectral parameters in HITRAN database: A novel SCLB model for predicting high-temperature gas spectra

Recent advancements in aerospace optics have highlighted the significance of high-temperature gas optical ra-diation properties in the fields of optical research and optical detection. The High-Resolution Transmission Molecular Absorption Database (HITRAN) provides a comprehensive range of high-temperature gas spectral radiation parameters, which are essential for studying the optical radiation of these gases. However, the coverage of common gas spectral line parameters in the HITRAN database is insufficient, particularly in supporting the multi-spectral research in the burgeoning fields of aerospace optics and optical detection, especially in the ul-traviolet range. This underscores the necessity of expanding the wavelength range of gas spectral parameters covered in the HITRAN database. In this study, we propose a physics-constrained LSTM-BPTransformer (SCLB) model that utilizes the distribution characteristics of gas spectral parameters in the database to predict spectral parameters for unknown wavelengths. The model combines data distribution quantification and distribution matching modules and incorporates the physical mechanisms of gas spectral radiation, thus enhancing prediction accuracy. Experimental results validate the effectiveness of the SCLB model, demonstrating that it can extend the data volume of gas spectral parameters in the HITRAN database by 4-5 times, while maintaining an error range within 1%. This predictive model enriches the research in high-temperature gas optics and lays a solid foun-dation for the application of high-temperature gases in the fields of optical research and optical detection.